S1E37: Marc Paradis, Northwell Holdings Commercializing AI data strategy for Health Systems

Share:
Marc Paradis, Vice President, Data Strategy at Northwell Holdings discusses the gap between AI data strategy and revenue generation.

Transcript

0:0:0.0 –> 0:0:5.70
Jordan Cooper
We’re here today with Mark Paradis, the vice president of data strategy at Northwell Holdings.

0:0:5.660 –> 0:0:6.910
Jordan Cooper
Mark, thank you for joining us today.

0:0:7.660 –> 0:0:8.690
Paradis, Marc
Thank you for having me, Jordan.

0:0:9.320 –> 0:0:18.980
Jordan Cooper
So for those who don’t know, Northwell Health is a health system based in Long Island, New York, with 5200 beds and 17,000 providers across 20 hospitals and other care facilities.

0:0:19.340 –> 0:0:25.350
Jordan Cooper
Now N will holdings to for profit arm of Northwell Health that focuses on commercialization of internal innovation.

0:0:25.880 –> 0:0:31.550
Jordan Cooper
Today I’d like to focus on bridging the gap between AI data strategy and revenue generation.

0:0:31.560 –> 0:0:45.520
Jordan Cooper
So Mark, I’d like to ask if you could walk us through an example of a time when someone approached you with an innovative data model workflow or something else being implemented at Northwell Health and how you evaluated that idea and ushered it to a point where it began to generate revenue?

0:0:46.790 –> 0:0:47.140
Paradis, Marc
Yeah.

0:0:47.150 –> 0:0:47.950
Paradis, Marc
No, thanks, Jordan.

0:0:47.960 –> 0:0:48.880
Paradis, Marc
It’s a. It’s a.

0:0:48.950 –> 0:0:50.80
Paradis, Marc
It’s a great question, right?

0:0:50.90 –> 0:0:57.210
Paradis, Marc
And it’s it’s a rather sort of expansive one and it’s worth, I think, giving a little bit of context to in terms of how to think about it.

0:0:57.410 –> 0:1:9.10
Paradis, Marc
So one of the important things, because ideas are coming all the time, right where a large health system, as you said 8485 thousand plus employees, you know all the time innovation is going on all the time.

0:1:9.20 –> 0:1:9.930
Paradis, Marc
People are having ideas.

0:1:9.940 –> 0:1:19.400
Paradis, Marc
We’re awash in data, and so one of the things it becomes very important is what is your framework for evaluating or thinking through these types of opportunities?

0:1:20.260 –> 0:1:36.230
Paradis, Marc
One of the things that that a mistakes that people often make is they look at the idea in isolation or they look at the idea as it came to the to the person or the individual and really what you need to do is take more of a product focus, right or a product management or product lifecycle type of view.

0:1:36.240 –> 0:1:45.490
Paradis, Marc
And so one of the things that becomes very important there is when these ideas come to you is you have to start asking questions, not necessarily about how is that idea being implemented now.

0:1:45.500 –> 0:1:46.590
Paradis, Marc
Do we have the data?

0:1:46.600 –> 0:1:51.570
Paradis, Marc
Is the data structured but you also have to start asking questions around who is going to be using this data.

0:1:51.640 –> 0:1:53.190
Paradis, Marc
Where in a workflow is this?

0:1:53.200 –> 0:1:55.230
Paradis, Marc
Uh, is this this new idea?

0:1:55.240 –> 0:1:57.10
Paradis, Marc
This application going to be inserted.

0:1:57.200 –> 0:2:2.110
Paradis, Marc
How is it going to change someone’s decision in order to hopefully lead to better outcomes?

0:2:2.260 –> 0:2:4.870
Paradis, Marc
And importantly, it’s not just about outcomes.

0:2:4.880 –> 0:2:9.20
Paradis, Marc
It’s also about the value that’s gonna be derived from that change in an outcome.

0:2:9.30 –> 0:2:12.970
Paradis, Marc
So you have to understand, are we looking primarily for a clinical outcome?

0:2:12.980 –> 0:2:15.490
Paradis, Marc
Are we looking primarily for a financial outcome?

0:2:15.800 –> 0:2:24.780
Paradis, Marc
Are we looking primarily for some type of a social or mission oriented outcome and all of that informs whether or not it’s a it’s a good idea.

0:2:24.850 –> 0:2:29.750
Paradis, Marc
I can’t tell you the number of times an idea comes to us, and it’s a great idea in the moment.

0:2:30.0 –> 0:2:43.250
Paradis, Marc
Uh, in a given workflow for a given individual or on a given unit, but when you begin to think through, is this something of value beyond just this one use case?

0:2:43.260 –> 0:2:50.780
Paradis, Marc
Is this something of value beyond just sort of the immediate need or opportunity that this individual noticed?

0:2:52.400 –> 0:2:56.550
Paradis, Marc
You realize that a lot of these things, a lot of these ideas don’t have the full package, right?

0:2:56.560 –> 0:2:57.110
Paradis, Marc
They’re not.

0:2:57.120 –> 0:2:57.940
Paradis, Marc
They’re not there yet.

0:2:57.950 –> 0:3:17.70
Paradis, Marc
They don’t have all of those parts and pieces, so as an example of something that did have all of those parts and pieces, as we do have an algorithm that we’ve developed that can go and take a variety of different inputs, both from the EMR as well as from our registration systems and previous history.

0:3:17.400 –> 0:3:28.590
Paradis, Marc
And it will go and predict with the reason we high degree of accuracy those individuals, uh, who are likely to have or likely to report a bad experience right during their stay.

0:3:28.800 –> 0:3:35.630
Paradis, Marc
And this is very important in healthcare from stars rating and patient satisfaction experience and all the rest of it.

0:3:35.990 –> 0:3:53.900
Paradis, Marc
And so being able to identify these individuals and then step in and do service recovery right, try to address what are going to be their issues before they become issues allows us to have a better outcome, right, and allows us to engage with those individuals in a way that they wish to be engaged with during their during their stay.

0:3:53.990 –> 0:4:1.900
Paradis, Marc
Now in that particular case, it’s relatively easy to go and see as you start to think down the line.

0:4:2.750 –> 0:4:3.940
Paradis, Marc
Is this worth looking at?

0:4:3.950 –> 0:4:5.30
Paradis, Marc
Is it worth building out?

0:4:6.500 –> 0:4:10.450
Paradis, Marc
You can start to think about, well, where and how is it gonna be inserted into workflow you.

0:4:10.460 –> 0:4:16.350
Paradis, Marc
While you can insert it shortly after an individual is is admitted to the hospital, you can run these sorts of predictive algorithms.

0:4:16.540 –> 0:4:19.870
Paradis, Marc
You can continue to update them during their stay as different events happen.

0:4:20.340 –> 0:4:27.810
Paradis, Marc
Once you identify that individual, it’s easy to engage in that service recovery effort when you have some idea or some sense of what that issue might be.

0:4:28.0 –> 0:4:34.310
Paradis, Marc
And then the outcome on the other end is hopefully an approved customer experience and improved patient experience.

0:4:35.230 –> 0:4:40.940
Paradis, Marc
Uh, and that that’ll show up in the in the header scores and otherwise.

0:4:40.950 –> 0:4:45.540
Paradis, Marc
Right, which has an impact both from a quality standpoint and from a from a financial standpoint.

0:4:45.550 –> 0:4:50.740
Paradis, Marc
And then the last piece of that as well is that’s a problem that’s not just a Northwell problem, right?

0:4:50.790 –> 0:4:56.880
Paradis, Marc
Every health system needs to understand which of their customers, which of their patients may or may not.

0:4:57.720 –> 0:5:8.860
Paradis, Marc
Uh, the having the experience that they deserve while they’re in the system, so that sort of one of these examples which was brought to us where we thought of it through end to end.

0:5:9.130 –> 0:5:16.510
Paradis, Marc
And in that context, it made sense for us to pursue and to create an actual application around and to commercialize that.

0:5:17.470 –> 0:5:23.860
Jordan Cooper
It’s an interesting example where you basically talk about how the squeaky wheel gets the grease before it squeaks.

0:5:23.950 –> 0:5:25.480
Jordan Cooper
I like that idea.

0:5:25.830 –> 0:5:31.60
Jordan Cooper
How have you been able to commercialize that idea, in particular outside of Northwell?

0:5:31.70 –> 0:5:38.290
Jordan Cooper
You mentioned that every health system needs to identify needs to report on HEDIS scores and predict you know who might be the most.

0:5:38.390 –> 0:5:39.930
Jordan Cooper
That the loudest complainer?

0:5:40.490 –> 0:5:43.450
Jordan Cooper
How do you commercialize it external to the organization?

0:5:44.520 –> 0:5:46.370
Paradis, Marc
So that’s a great question as well.

0:5:46.380 –> 0:5:47.510
Paradis, Marc
And there are.

0:5:47.760 –> 0:5:51.990
Paradis, Marc
So I’ll tell you specifically with this algorithm we chose to.

0:5:52.420 –> 0:5:59.50
Paradis, Marc
We chose to commercialize it externally by licensing it to a third party and that third party already has a suite of.

0:6:0.60 –> 0:6:7.420
Paradis, Marc
A algorithms and and applications around customer experience and managing customer experience and helping to do service recovery.

0:6:7.650 –> 0:6:10.930
Paradis, Marc
And so this is another sort of interesting decision, right?

0:6:10.940 –> 0:6:17.320
Paradis, Marc
One of the decisions that you have to make when you build these algorithms is are you going to engage?

0:6:17.330 –> 0:6:17.870
Paradis, Marc
Are you sort of?

0:6:17.880 –> 0:6:22.440
Paradis, Marc
Are you going to become, in effect, almost a software development company, right?

0:6:22.450 –> 0:6:25.750
Paradis, Marc
And where you’re going to go out, you’re going to build the application, you’re going to provide the support.

0:6:25.760 –> 0:6:33.230
Paradis, Marc
You’re going to take that out in the market, are you then going to also go to the capital markets, right, to raise funds around that and support that?

0:6:33.370 –> 0:6:36.380
Paradis, Marc
And for some of the ideas, it makes sense for us to do that, right?

0:6:36.390 –> 0:6:52.720
Paradis, Marc
It makes sense for us to keep that type of equity because the and that type of ownership, because the idea has the appropriate potential or there are enough A other algorithms or or IP that we can add to that concept.

0:6:53.200 –> 0:6:56.460
Paradis, Marc
But in this case, when we looked at it, this was kind of a standalone algorithm.

0:6:56.470 –> 0:6:59.290
Paradis, Marc
We didn’t really have a suite that we wanted to build around it.

0:6:59.300 –> 0:7:6.860
Paradis, Marc
And when we looked at the market, there were already companies out there that did a pretty good job of of helping health systems, right.

0:7:6.870 –> 0:7:10.600
Paradis, Marc
Do this type of of service recovery and and manage that experience.

0:7:10.610 –> 0:7:14.510
Paradis, Marc
And so we licensed it to a third party in there.

0:7:15.920 –> 0:7:17.760
Paradis, Marc
They’re they’re deployed it as part of their package.

0:7:18.740 –> 0:7:25.150
Jordan Cooper
Now I also know that as part of Northwell Holdings you manage a portfolio of companies.

0:7:25.860 –> 0:7:33.810
Jordan Cooper
I know that many different health systems listening to this episode right now have different accelerator incubator programs.

0:7:33.920 –> 0:7:42.570
Jordan Cooper
I know that you manage Holdings manages direct ventures, joint ventures, data partnerships in addition to commercializing innovation from Northwell Health.

0:7:42.760 –> 0:7:50.140
Jordan Cooper
Would you speak about actually owning equity in these portfolio companies or or how the accelerator process works?

0:7:50.230 –> 0:7:50.760
Jordan Cooper
What?

0:7:50.930 –> 0:7:53.890
Jordan Cooper
You know what’s going on in in that in that sphere.

0:7:55.290 –> 0:7:55.720
Paradis, Marc
Yeah.

0:7:55.730 –> 0:7:58.940
Paradis, Marc
So we’ve actually taken a little bit of a a different approach.

0:7:59.650 –> 0:8:24.560
Paradis, Marc
There are look I I think everyone is aware of the pressures that health systems are under, right, not only from from recent inflationary pressures and and labor costs and COVID the sort of lingering effects of COVID and the withdrawal of some of the COVID funds, but even just if you look at long term trends, the health systems are all working on supermarket margins that are increasingly being squeezed.

0:8:25.340 –> 0:8:38.110
Paradis, Marc
The growth in the market is you know sort of in the Medicaid Medicare side of the world, right, which is which is not where I am, not where the high margin of work is.

0:8:38.120 –> 0:8:48.490
Paradis, Marc
And so every health system needs to find ways to diversify their revenue, or they’re gonna face some serious financial crunches in the next 5 to 10 years, UM.

0:8:48.580 –> 0:8:58.550
Paradis, Marc
And so when we looked at that, certainly one of the things that people do is they create these corporate venture capital arms where they go out and they raise a fund from a variety of investors.

0:8:58.560 –> 0:9:5.610
Paradis, Marc
They go out and they look for they leverage their healthcare expertise, they look for some smart healthcare investments, right, great strategy, a lot of people do that.

0:9:6.300 –> 0:9:14.950
Paradis, Marc
There’s another strategy that a lot of people use, which is, as you mentioned, this notion of the accelerator, where they’ll go and they’ll find a couple of early stage companies.

0:9:14.960 –> 0:9:19.60
Paradis, Marc
They’ll do some type of a Shark Tank like type of event.

0:9:19.230 –> 0:9:20.620
Paradis, Marc
They’ll pick a couple of winners.

0:9:20.630 –> 0:9:23.180
Paradis, Marc
They’ll work closely with those winners over a period of time.

0:9:23.190 –> 0:9:35.760
Paradis, Marc
Give them access to resources and data and expertise, and then sort of gently nudge them out of the nest on the other side and maintain some level of equity or ownership in return for that.

0:9:36.90 –> 0:10:8.290
Paradis, Marc
Those are all fantastic models, but we really sort of looked at it as we wanted to make sure that one that the investments that we made were very directly tied to tied to mission and that that the the overriding goal of what we’re doing wasn’t necessarily just to maximize the multiples but to generate multiples that supported our mission and were in areas that we felt were under invested in or underrepresented.

0:10:9.40 –> 0:10:19.410
Paradis, Marc
So we’ve got $10 million annual off balance sheet fund that we invest in a variety of companies, again around things like diversity, equity and inclusion.

0:10:20.580 –> 0:10:31.790
Paradis, Marc
Around helping to make the data and the medical record more understandable and more readable, helping to reduce A barriers to access.

0:10:31.980 –> 0:10:41.870
Paradis, Marc
Whether those barriers are with respect to to transport or social determinants of health, or working through the paperwork or any variety of other things.

0:10:41.940 –> 0:10:43.450
Paradis, Marc
So that’s part of what we do.

0:10:43.460 –> 0:10:53.560
Paradis, Marc
But then the other piece of it, which I think is a little bit different, is we’ve actually formed this partnership with a startup studio called Egis Ventures.

0:10:53.790 –> 0:11:3.980
Paradis, Marc
It’s a this particular partnership is a little bit over a year old and we are working together sort of bringing together, I guess maybe the chocolate and the peanut butter, if you will.

0:11:3.990 –> 0:11:13.160
Paradis, Marc
Are there a piece of these two ideas where you just ventures brings the speed and the agility of a startup studio?

0:11:13.170 –> 0:11:20.960
Paradis, Marc
They bring access to capital markets, they bring expertise and commercialization and building startups and getting those steps up and running off the ground.

0:11:21.390 –> 0:11:21.600
Jordan Cooper
Mm-hmm.

0:11:21.270 –> 0:11:44.870
Paradis, Marc
Whereas Northwell on the other side brings their healthcare expertise, their operational expertise, right, the healthcare delivery platform that we have here and we really bring those two things together where we go and as we sort of start off that question or you’re very first question there, we identify these problems or challenges at Northwell we ask are these just northwell problems or are they broader problems?

0:11:44.880 –> 0:11:51.700
Paradis, Marc
If they’re broader problems, we run them through that very sort of strict product management, not product life cycle, product focus, type of view.

0:11:51.960 –> 0:12:9.790
Paradis, Marc
And if there’s an opportunity there and it makes sense, then we bring together all of that into to a, you know, transparent, repeatable evidence based, methodologically rigorous pipeline that we run these ideas through from beginning all the way out to the end when they get launched into the market.

0:12:9.800 –> 0:12:25.380
Paradis, Marc
As companies go for a seed or a series, A uh and ultimately uh putting products out into the market that are that are effective and impactful based on based on AI and data in the background.

0:12:25.970 –> 0:12:42.650
Jordan Cooper
So it almost sounds like Northwell Health by operating Northwell Holdings or having the the relationship with Northwell Holdings has kind of diversified its financial portfolio from providing clinical care to almost entering the private equity space.

0:12:55.260 –> 0:12:55.680
Jordan Cooper
Ohh.

0:12:44.240 –> 0:13:1.20
Paradis, Marc
Uh, so I wouldn’t necessarily call it the private equity space because private equity depending on how you feel about private equity, they get some bad press, right, either either they’re either they’re the engine that drives the economy or they or they ruin, they ruin the economy.

0:13:1.30 –> 0:13:11.100
Paradis, Marc
So I wouldn’t necessarily frame it as private equity, but what I would phrase it as is more this notion of really trying to take the best of these two different worlds, right?

0:13:11.110 –> 0:13:14.940
Paradis, Marc
So look at the end of the day, we’re a healthcare delivery system.

0:13:14.950 –> 0:13:15.990
Paradis, Marc
We are about patients.

0:13:16.0 –> 0:13:17.500
Paradis, Marc
We’re about delivering world class.

0:13:17.510 –> 0:13:18.550
Paradis, Marc
Excellent care.

0:13:18.700 –> 0:13:19.820
Paradis, Marc
We’re about outcomes.

0:13:20.190 –> 0:13:26.400
Paradis, Marc
Importantly, we’re also about the administrative back end behind that because that’s part of healthcare, right?

0:13:26.410 –> 0:13:27.180
Paradis, Marc
There is no healthcare.

0:13:33.700 –> 0:13:33.890
Jordan Cooper
Mm-hmm.

0:13:27.190 –> 0:13:34.710
Paradis, Marc
If you don’t get paid and we happen to have the system that we have as a sort of a mix of private and public payers, right, and Medicare, Medicaid and all that.

0:13:34.720 –> 0:13:48.130
Paradis, Marc
So we’ve got it very deep expertise in in all of that back end and how that how that works A and also obviously a lot of expertise in the actual sort of technology of of healthcare, right.

0:13:48.140 –> 0:13:51.780
Paradis, Marc
And so that is what generates the ideas.

0:14:0.810 –> 0:14:1.20
Jordan Cooper
Mm-hmm.

0:13:51.790 –> 0:14:25.720
Paradis, Marc
That is what gives the platform to to to to test out those ideas and demonstrate in a in a safe, rigorous way, right, the positive impact of that and then really sort of tying that up to again this the that startup studio, that sort of venture capital type of idea where you’re moving UMA moving at a speed that allows you to actually get impactful ideas out into the market, to lead to better outcomes.

0:14:26.620 –> 0:14:32.800
Jordan Cooper
So Mark the focus of today’s conversation is the gap between AI data strategy and revenue generation.

0:14:32.980 –> 0:14:33.120
Paradis, Marc
Yeah.

0:14:35.700 –> 0:14:35.840
Paradis, Marc
Yeah.

0:14:32.810 –> 0:14:40.830
Jordan Cooper
And we’ve covered the second one to a great extent, revenue generating generation commercialization, I’d like to return us to a I data strategy.

0:14:41.280 –> 0:14:42.410
Jordan Cooper
What’s going on?

0:14:42.620 –> 0:14:43.550
Jordan Cooper
Where has been.

0:14:43.560 –> 0:15:0.0
Jordan Cooper
I know that we’ve spoken previously and you said that you’re looking to reduce a cognitive workload on physicians and that a lot of issues with AI or with EHR or clinical decision support is that it requires more physician attention detracting from the patient experience.

0:15:0.180 –> 0:15:0.340
Paradis, Marc
Yeah.

0:15:0.510 –> 0:15:9.700
Jordan Cooper
Can you speak to some different trends that you’re seeing emerge in the and general AI data strategy that’s driving decisions at Northwell?

0:15:10.230 –> 0:15:11.850
Paradis, Marc
Yeah, absolutely.

0:15:12.330 –> 0:15:13.130
Paradis, Marc
Well, that’s the.

0:15:14.890 –> 0:15:17.960
Paradis, Marc
You know the the elephant or the 800 pound gorilla in the room, right?

0:15:20.860 –> 0:15:21.60
Jordan Cooper
Mm-hmm.

0:15:17.970 –> 0:15:21.100
Paradis, Marc
That everybody’s talking about these days is generative AI, of course, right?

0:15:21.110 –> 0:15:29.250
Paradis, Marc
So large language models and all of the work that’s going on there, and that’s certainly an area that we are looking at very closely and.

0:15:30.520 –> 0:15:41.460
Paradis, Marc
Figuring out where and how to place our bets from an investment standpoint to really impact care and coming back to sort of where we started from or where you started from in this particular question.

0:15:42.0 –> 0:15:54.120
Paradis, Marc
So much of a clinicians day, uh is is just loaded with A tasks and interactions and decision making.

0:15:54.130 –> 0:15:56.600
Paradis, Marc
Oftentimes very, very, very high stakes decision making.

0:15:56.610 –> 0:16:9.530
Paradis, Marc
Oftentimes, on the basis of incomplete information, right to go and try and insert an extra step into that workflow, even something as simple as a drop down, it breaks up that that thinking right, it slows things down.

0:16:9.540 –> 0:16:11.690
Paradis, Marc
Ask them to make one or two extra clicks.

0:16:12.120 –> 0:16:13.850
Paradis, Marc
Again, very, very problematic.

0:16:13.860 –> 0:16:21.510
Paradis, Marc
It seems so simple, but I would challenge anyone to spend a couple of hours in a busy emergency department right.

0:16:21.520 –> 0:16:28.680
Paradis, Marc
Shadowing, shadowing and Ed Doc or spend some time in a surgical suite right during one of these complex surgeries.

0:16:28.690 –> 0:16:35.640
Paradis, Marc
And I think you’ll get some sense for the importance of having those clinicians have that level of focus, right.

0:16:35.650 –> 0:16:38.860
Paradis, Marc
And letting those clinicians have that level of attention.

0:16:39.330 –> 0:17:13.570
Paradis, Marc
And so that’s has been one of our driving factors from the very beginning when we’re addressing these ideas or thinking about them again, not just looking at the idea in isolation, but thinking about once this goes into production once it’s in and application, where does it sit in the workflow from very early on, we’re talking to the clinicians about if you had an application that did this A where and when would you want that if you had an application where the the algorithm worked perfectly in terms of it’s predictability where and when in the workflow would you want to know that piece of information and how would it help you to make it different decision from that standpoint.

0:17:13.660 –> 0:17:22.170
Paradis, Marc
So coming back to the sort of the AI strategy piece of it, I do want to make another point that I think is worth talking about.

0:17:23.550 –> 0:17:26.350
Paradis, Marc
A lot of people think of data strategy as one thing.

0:17:26.430 –> 0:17:29.480
Paradis, Marc
They think of AI strategy as something else, right?

0:17:29.490 –> 0:17:33.930
Paradis, Marc
They may sort of think of your digital transformation, your digital strategy as a third thing, right.

0:17:33.940 –> 0:17:36.680
Paradis, Marc
And then maybe you’ve got a commercialization strategy out here.

0:17:36.690 –> 0:17:46.390
Paradis, Marc
And I do think that’s one of the having those silos makes the transitions between those silos very, very difficult and very, very inefficient.

0:17:46.400 –> 0:17:52.740
Paradis, Marc
And I’ve seen a lot of good ideas fail because they worked in one silo but couldn’t survive the transition to the next.

0:17:52.750 –> 0:18:12.110
Paradis, Marc
So as we think about it, when we’re talking about data strategy, that data strategy is inclusive not only of all the sort of data governance, chief Chief data officer types of things you’d be thinking about, but it’s also inclusive of the algorithms that you are building and validating and deriving off of that data.

0:18:12.120 –> 0:18:13.490
Paradis, Marc
And then it is also inclusive.

0:18:13.540 –> 0:18:19.670
Paradis, Marc
Ultimately, of the applications that you’re embedding those algorithms into to drive these changes in behavior, right?

0:18:19.680 –> 0:18:23.630
Paradis, Marc
Hopefully being to better outcomes and net new value.

0:18:23.640 –> 0:18:32.210
Paradis, Marc
So from that standpoint, the conversational abilities of these LLNS are really intriguing, right?

0:18:32.220 –> 0:18:48.360
Paradis, Marc
Because certainly in a surgical suite, right, all of the clinicians are talking to each other already and interacting in that in that busy Ed the all again all of the clinicians are interacting with each other and talking to each other, oftentimes with the patient and their family in the room as well.

0:18:48.610 –> 0:18:56.30
Paradis, Marc
And so having these conversational interfaces, which can then be connected to effectively the world’s data store.

0:18:56.40 –> 0:18:56.260
Paradis, Marc
Right.

0:18:56.270 –> 0:18:57.800
Paradis, Marc
Or the world’s knowledge.

0:18:58.450 –> 0:19:15.510
Paradis, Marc
In an ideal world, it would be like having the best, smartest, most up-to-date resident you’ve ever worked with, just sort of standing there behind your shoulder to have a conversation with, to consult with, to ask a quick question, or to maybe make a point out.

0:19:15.520 –> 0:19:21.790
Paradis, Marc
A small detail in one place or another, and that can be done in a way that is non intrusive, right?

0:19:21.800 –> 0:19:34.120
Paradis, Marc
And that allows the clinicians to still practice allows the clinicians to engage with the patients in ways that a lot of the EHR and a lot of these other apps are currently pulling them away from at the moment.

0:19:34.610 –> 0:19:43.700
Jordan Cooper
So just to clarify, you’re saying that there’s some discussion about combining large language models with natural language processing, kind of like Amazon’s Alexa.

0:19:43.850 –> 0:19:45.890
Jordan Cooper
So you can speak to them to the actual.

0:19:47.370 –> 0:19:47.730
Jordan Cooper
Chat.

0:19:47.740 –> 0:19:48.680
Jordan Cooper
GBT.

0:19:49.20 –> 0:19:50.430
Jordan Cooper
Verbally, no need to type.

0:19:50.440 –> 0:19:51.180
Jordan Cooper
Is that what you’re saying?

0:19:51.470 –> 0:19:52.270
Paradis, Marc
Exactly right.

0:19:52.280 –> 0:19:53.230
Paradis, Marc
And it can do the same back.

0:19:53.240 –> 0:19:55.100
Paradis, Marc
You know, it can also speak to you as well, right?

0:19:55.110 –> 0:20:0.600
Paradis, Marc
So it does not only you talking to the to that to that large language model, it’s also the large language model.

0:20:0.610 –> 0:20:8.190
Paradis, Marc
Speaking back to you, right, being able to have that very as humans the 1st way we learned to communicate right was with our voices.

0:20:8.250 –> 0:20:8.350
Paradis, Marc
This.

0:20:8.620 –> 0:20:14.370
Paradis, Marc
Yeah, well, maybe the first one was maybe with our hands right by pointing but shortly thereafter right. We we had.

0:20:19.590 –> 0:20:19.770
Jordan Cooper
Mm-hmm.

0:20:14.380 –> 0:20:20.490
Paradis, Marc
We had voices in verbal and it’s just a much more natural form of communication for us than typing.

0:20:20.660 –> 0:20:20.940
Paradis, Marc
Right.

0:20:22.260 –> 0:20:35.490
Paradis, Marc
And and so from that standpoint it’s it’s just kind of baked into our our biology to be able to have a conversation while still be able to being able to concentrate on a task to be able to collaborate via language.

0:20:35.660 –> 0:20:51.780
Paradis, Marc
And so that’s really one of the areas we’re looking at is how can we leverage these large language models and transformer networks and generative AI in general to really begin to engage with us in ways that are natural and seamless. Right.

0:20:51.960 –> 0:20:54.330
Paradis, Marc
Uh, rather than intrusive, right?

0:20:54.340 –> 0:20:59.40
Paradis, Marc
And ways that that keep us in the in the work rather than pulling us out of the work.

0:20:59.690 –> 0:21:3.160
Jordan Cooper
Well, Mark, I think we’ve come to the end of our podcast episode.

0:21:3.170 –> 0:21:8.520
Jordan Cooper
It’s been an exciting conversation about the topic of AI, data strategy and revenue generation.

0:21:8.830 –> 0:21:17.60
Jordan Cooper
We have spoken about algorithms that have made physician workflows more efficient and also non position clinical workflows.

0:21:17.270 –> 0:21:27.840
Jordan Cooper
We’ve spoken about diversifying revenue streams for health systems and by partnering with venture capital and startup entities.

0:21:28.70 –> 0:21:35.790
Jordan Cooper
And then we’ve also spoken about the future of generative AI, combining LM’s with NLP.

0:21:35.920 –> 0:21:42.140
Jordan Cooper
So for our listeners, this has been Mark Paradis, the vice president of data strategy at Northwell Holdings Market.

0:21:42.150 –> 0:21:43.800
Jordan Cooper
Like to thank you so much for joining us today.

0:21:44.530 –> 0:21:44.940
Paradis, Marc
Thank you.

0:21:44.950 –> 0:21:46.150
Paradis, Marc
Thank you very much for having me, Jordan.